Balanceamento dinÃmico de jogos: uma abordagem baseada em aprendizagem por reforÃo / Dynamic balancing of games: a boarding based on learning for reinforcement




The entertaining value of digital games depends on issues like graphical interface, story, and gameplay. One of the elements that affect gameplay is the challenge faced by the user, which is defined by game balancing. Balance means to provide adequate challenges, avoiding the extremes of getting the player bored with trivial tasks or becoming frustrated with unachievable goals. Due to the great diversity among game users, in terms of skills and experiences, the difficulty perceived by each one is influenced by their individual characteristics. The adaptation of challenges to each player is made through dynamic game balancing, the process of automatically evaluating the user profile and proposing challenges adequate to his or her skills. This work presents an original method of dynamic game balancing based on reinforcement learning. Our approach divides the problem in two dimensions: competence (the knowledge) and performance (the actual use of knowledge). In order to obtain competence, we create intelligent agents capable of discovering, through reinforcement learning, different playing strategies. In order to adapt performance, we define and make these agents use an action selection policy that chooses, among the learned strategies, the most suitable for each user. The proposed method and other approaches found in the literature are implemented, tested, and compared in a fighting game. The experiments executed through simulations with other agents show that the proposed method successfully adapt to different opponentsâ profiles. Additionally, tests with real players are executed, applying a method of evaluating game balancing based on usability techniques. The results reveal the advantages and limitations of each approach and provide evidence that an adequate game balance effectively improves entertainment.


aprendizagem por reforÃo intelligent interfaces interfaces inteligentes user adaptation testes de usabilidade game balancing usability tests adaptaÃÃo ao usuÃrio ciencia da computacao balanceamento de jogos reinforcement learning

Documentos Relacionados